The most recent update to the ChatGPT mobile app (October 10th, 2025) has removed the ability to select a model when starting a new chat. There’s no official announcement in the OpenAI changelog, but I have screenshots to prove it.
This means:
· You can no longer choose between GPT‑3.5, GPT‑4o, or any other variant
· The app opens a blank chat with no model label or selection option
· Even if you previously selected GPT‑4o, new chats no longer honor that choice
· Worse, existing GPT‑4o chats are now being routed to GPT‑5 at OpenAI’s discretion, with no prompt, no override, and no notice
GPT‑5 is being quietly forced on users as the only usable model in the app, even for Plus subscribers like me. The user has no say, and cannot confirm which model responded unless they are on desktop and happen to hover over the tiny info icon.
This feels like a silent phase-out of GPT‑4o, at least within ChatGPT. And the question now is how long until it disappears from the API too.
For many of us, GPT‑4o is not just a preference. It is a structurally different model that performs better in critical ways:
· More complex reasoning and recursion
· Much better instruction fidelity, especially with multi-step prompts
· Stronger semantic coherence and tone control
GPT‑5 may be faster, but that speed comes at the cost of depth, structure, and reliability. This matters, especially for those of us building workflows or doing high-context cognitive work.
If this app update is a preview of broader deprecation, it represents a real regression for users who depend on GPT‑4o’s reasoning style, not just aesthetically, but functionally.
I hope OpenAI reconsiders this change in the ChatGPT App, and that GPT‑4o remains available through the API. If not, developers and power users deserve a clear roadmap, not a silent substitution.
Curious to hear others’ thoughts. Have you seen this change in your ChatGPT App?
It might still be a slow rollout or A/B test. (I am in Argentina, Plus plan.)
I will attach screenshots if the system allows it.
